Scholarlink

Emerging research in affective computing and intelligent environments highlights the need for adaptive thermal comfort systems that address individualized discomfort in real-time within educational settings. Traditional HVAC systems cannot dynamically respond to student-specific needs, which can reduce engagement and learning outcomes. To address this gap, this paper presents the design and evaluation of an IoT-based Smart Desk system that integrates low-power hardware and multimodal sensor fusion for desk-level comfort classification. Each desk incorporates an ESP32-WROVER module with an MLX90640 infrared thermal array to detect facial skin temperature and an onboard camera to capture facial images. A central ESP32 node monitors the ambient temperature and humidity through a DHT11 sensor. Data from all desks are transmitted to an instructor s computer, where facial emotion recognition is performed using DeepFace with pre-trained convolutional neural networks (CNN), such as VGG-Face and FaceNet. Expressions are classified as happy, neutral, sad, and angry. These emotion labels, combined with the readings of skin temperature and ambient sensor, are evaluated against the ASHRAE-guided thresholds, Ambient temperature: 19– \(27\,^{\circ }\textrm{C}\) (66– \(80\,^{\circ }\textrm{F}\) ), relative humidity: 30– \(60\,\%\) , skin temperature: 32.6– \(33.5\,^{\circ }\textrm{C}\) ( \(\approx 90\) – \(92\,^{\circ }\textrm{F}\) ), and emotion \(\in \{\text {happy},\ \text {neutral}\}\) . Rule: if \(\ge \tfrac{3}{4}\) conditions hold \(\rightarrow \) Comfortable; else Uncomfortable, and the desk labels are aggregated into a room-level comfort score. The proposed system is optimized for privacy, bandwidth, and power, making it suitable for scalable deployment in real-world classrooms. By combining affective state recognition with environmental sensing, the Smart Desk platform bridges the gap between individual comfort detection and classroom-wide adaptive environmental control. Future work will focus on large-scale validation, predictive comfort modeling, and closed-loop HVAC integration through Google Nest APIs, enabling proactive and automated classroom climate management.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Multimodal IoT-Based Smart Desk System for Real-Time Thermal Comfort Classification in Educational Environments

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Krupa V. Khapper,
  • Issa W. AlHmoud,
  • Balakrishna Gokaraju,
  • AKM Kamrul Islam,
  • Corey A Graves

摘要

Emerging research in affective computing and intelligent environments highlights the need for adaptive thermal comfort systems that address individualized discomfort in real-time within educational settings. Traditional HVAC systems cannot dynamically respond to student-specific needs, which can reduce engagement and learning outcomes. To address this gap, this paper presents the design and evaluation of an IoT-based Smart Desk system that integrates low-power hardware and multimodal sensor fusion for desk-level comfort classification. Each desk incorporates an ESP32-WROVER module with an MLX90640 infrared thermal array to detect facial skin temperature and an onboard camera to capture facial images. A central ESP32 node monitors the ambient temperature and humidity through a DHT11 sensor. Data from all desks are transmitted to an instructor s computer, where facial emotion recognition is performed using DeepFace with pre-trained convolutional neural networks (CNN), such as VGG-Face and FaceNet. Expressions are classified as happy, neutral, sad, and angry. These emotion labels, combined with the readings of skin temperature and ambient sensor, are evaluated against the ASHRAE-guided thresholds, Ambient temperature: 19– \(27\,^{\circ }\textrm{C}\) (66– \(80\,^{\circ }\textrm{F}\) ), relative humidity: 30– \(60\,\%\) , skin temperature: 32.6– \(33.5\,^{\circ }\textrm{C}\) ( \(\approx 90\) – \(92\,^{\circ }\textrm{F}\) ), and emotion \(\in \{\text {happy},\ \text {neutral}\}\) . Rule: if \(\ge \tfrac{3}{4}\) conditions hold \(\rightarrow \) Comfortable; else Uncomfortable, and the desk labels are aggregated into a room-level comfort score. The proposed system is optimized for privacy, bandwidth, and power, making it suitable for scalable deployment in real-world classrooms. By combining affective state recognition with environmental sensing, the Smart Desk platform bridges the gap between individual comfort detection and classroom-wide adaptive environmental control. Future work will focus on large-scale validation, predictive comfort modeling, and closed-loop HVAC integration through Google Nest APIs, enabling proactive and automated classroom climate management.